DistFit (dsh-easydistfit): A DSH plugin for turning numerical tables into distribution fitting reports

DistFit is a DeepSeek Harness plugin maintained by stormx-u designed to transform numerical data into distribution fitting reports for scientific research scenarios. It fits 23 distribution families simultaneously and performs four types of goodness-of-fit tests (K-S, A-D, CvM, and χ²) with default asymptotic p-values, switchable Monte Carlo calibration. It ranks the distributions based on AIC/BIC/Akaike weights and provides conclusions of Acceptance, Partial Rejection, or Rejection. The graphics tab includes Histogram+PDF, CDF, P-P, and Q-Q plots, all rendered in vector SVG format, supporting hover-to-inspect values and legend point selection. After each calculation, the Host-side generates a four-paragraph AI interpretation using DSH's LLM service and the assistant's default model; the browser does not handle credentials. Chart export offers seven template sets (Chinese journal two-column, English journal single-column, thesis, 16:9, transparent background, black-and-white print, default), with adjustable font size, line width, grayscale, and bar styles. Previews are rendered off-screen at export resolution, supporting SVG, PNG, JPG, and PDF. There is also an "AI Adjustment" tab where a single sentence describing the intended use allows the model to generate a style patch subject to whitelist constraints. The statistical engine is entirely pure JavaScript, with no dependencies on Python or SciPy; it also registers the `distfit_fit` tool to allow the assistant to perform direct fitting within conversations. Installation is performed using `dsh plugin --profile web add link:"<repository path>"`, requiring DSH >= 0.1.7-rc.1 and Node.js 24+ for building.